[FFmpeg-devel] [PATCH 4/5] vaapi_encode: Add ROI support

Guo, Yejun yejun.guo at intel.com
Thu Mar 7 05:25:06 EET 2019



> -----Original Message-----
> From: ffmpeg-devel [mailto:ffmpeg-devel-bounces at ffmpeg.org] On Behalf
> Of Mark Thompson
> Sent: Tuesday, March 05, 2019 8:26 AM
> To: ffmpeg-devel at ffmpeg.org
> Subject: Re: [FFmpeg-devel] [PATCH 4/5] vaapi_encode: Add ROI support
> 
> On 28/02/2019 06:33, Guo, Yejun wrote:>> -----Original Message-----
> >> From: ffmpeg-devel [mailto:ffmpeg-devel-bounces at ffmpeg.org] On
> Behalf
> >> Of Mark Thompson
> >> Sent: Thursday, February 28, 2019 6:00 AM
> >> To: ffmpeg-devel at ffmpeg.org
> >> Subject: [FFmpeg-devel] [PATCH 4/5] vaapi_encode: Add ROI support
> >>
> >> ---
> >>  libavcodec/vaapi_encode.c       | 129
> >> ++++++++++++++++++++++++++++++++
> >>  libavcodec/vaapi_encode.h       |   9 +++
> >>  libavcodec/vaapi_encode_h264.c  |   2 +
> >>  libavcodec/vaapi_encode_h265.c  |   2 +
> >>  libavcodec/vaapi_encode_mpeg2.c |   2 +
> >>  libavcodec/vaapi_encode_vp8.c   |   2 +
> >>  libavcodec/vaapi_encode_vp9.c   |   2 +
> >>  7 files changed, 148 insertions(+)
> >
> > I tried this patch with below command, but do not see any quality change
> with my eyes.
> > I debugged in gdb and  found the ROI data are correct in function
> vaapi_encode_issue.
> >
> > I do not investigate deeper, and just want to first confirm if you see the
> quality change or not. I might did something basic wrong.
> >
> > ffmpeg_g -hwaccel vaapi -vaapi_device /dev/dri/renderD128  -s 1920x1080
> -i ../video
> > /1080P_park_joy_1920x1080_500frames.yuv   -vf format=nv12,hwupload
> -c:v h264_vaapi  -b:v 3M  -y tmp.264
> > (my trick code in vf_scale.c is called with the above command)
> >
> > I tried the similar option with libx264 and found obvious video quality
> changes.
> 
> If you are using the i965 driver then you might need
> <https://github.com/intel/intel-vaapi-driver/pull/447> to make it work.  The
> iHD driver worked for me with no changes.
> 
> In CQP mode with H.264 it's straightforward to verify the output directly, too
> - any stream analyser or other tool which can show the QP on each
> macroblock will make it very obvious, since you will see exactly the offset you
> set in your regions of interest.  (The reference decoder with trace enabled
> shows it as mb_qp_delta as well.)

yes, CQP mode is more straightforward. I can see the difference obviously with
my eyes when using option "-qp 50", and yes, with the analyzer tool, I can see
each MB's final qp is set as expected, but do not see the difference in 'slice_qp_delta'
and 'mb_qp_delta' in the analyzer.

> 
> Thanks,
> 
> - Mark
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
> https://ffmpeg.org/mailman/listinfo/ffmpeg-devel


More information about the ffmpeg-devel mailing list